[llvm] [ThinLTO] Don't mark calloc function dead (PR #72673)

[via llvm-commits]

eleviant wrote:

> Right, I don't think it is correct to special case calloc or any function. In general, this kind of functions are handled by preserving all RuntimeLibcalls

I don't think it's a good idea to treat all RuntimeLibcalls like this. The problem arises only when two (or more) libcalls are folded into one (or more) other libcalls.  This happens in particular when we import  operator new from another TU so DSE sees malloc followed by memset to zero. What are other cases possible? 

https://github.com/llvm/llvm-project/pull/72673